Theory and practice of Multiple-point Statistics with Isatis.neo | Training course

Multiple-point Statistics (MPS) is a geostatistical simulation technique based on training images. MPS can address many areas like hydrogeology, mining, remote sensing, or hydrocarbon reservoirs. It has its origins in geological facies modeling but can potentially be employed in any field requiring simulating complex spatial variability, categorical or continuous.

The course, developed in collaboration with the University of Neuchâtel, aims to familiarize you with the fundamental aspects of the Multiple-point Statistics approach and give you a hands-on experience through a series of practical exercises on a variety of case studies.

At the end of the course, you will be able to:
– select the appropriate training image according to your knowledge of the study area and the expected results,
– produce realistic subsurface models,
– prepare data and run MPS with Isatis.neo that integrates DeeSse, the advanced MPS code from the Swiss University of Neuchâtel.

• Half of the course is devoted to theoretical and methodological presentations, the second half to practical exercises on real-life cases to deepen the understanding of concepts based on real data (geology, mining). The focus is on illustrations and practical contributions of the covered concepts.
• Computer exercises with Isatis.neo Standard Edition.
• Course material provided (documentation, batch files, training data, worked examples) for re-use in your workplace.

This course aims at any scientist wishing to delve into MPS: academics, agricultural engineers, air quality engineers, climatologists, environmental consultants and engineers, epidemiologists, foresters, geologists, geophysicists, geotechnical engineers, hydrogeologists, hydrologists, mining resource specialists, reservoir engineers, soil scientists, etc.

DAY 1

MORNING

• General introduction
  – Overview of the geostatistical approach
  – The underlying concept of training data set and training image
  – General principle and introduction to the direct sampling algorithm
• Laboratory exercises
  – Fundamentals of Isatis.neo
  – A first simple application of DeeSse for a stationary case categorical and continuous

AFTERNOON

• From stationary to non-stationary simulation
  – Understanding DeeSse parameters
  – Requirement of a training image: how to get it and what should be its properties
  – Dealing with non-stationarity in the simulation grid
  – Multivariate simulations
• Laboratory exercises
  – A simple practical case study: the Areuse delta
  – How to generate a training image and a trend of orientation to control the simulation
  – Joint simulation of two variables

DAY 2

MORNING

• MPS using actual data for training
  – How to deal with non-stationarity when using analog data
  – Discussion of examples, the use of secondary attributes: climate data, bauxite mine in Australia, bedrock topography, and geophysics
  – Time-series simulation using the Direct Sampling technique
• Laboratory exercises
  – A 2D practical case study with secondary variables: the Herten aquifer (fluvioglacial deposit)
  – Multivariate, multitemporal satellite image gap-filling

AFTERNOON

• Modeling with elementary training images
  – Elementary training images and invariances
  – Example of application for a mining site in South Africa
  – Multiscale simulations with Gaussian Pyramids
• Laboratory exercises
  – Simple examples with elementary training images and invariances
  – Exploring pyramids
  – An initial example with a 2D fluvioglacial facies model (the Herten aquifer)

Day 3

MORNING

• Laboratory exercises: Modeling a fluvioglacial deposit
  – Construction of elementary training images
  – Introduction to Python scripting to automatize the tasks
  – Construction of the stratigraphical model
  – Modeling the fluvioglacial aquifer from borehole data

AFTERNOON

• A glimpse at advanced methods
  – Cross-validation
  – Multiscale simulations on unstructured grids
  – Inequality and block conditioning
  – Connectivity conditioning

None.
A theoretical knowledge of geostatistical approaches is a plus.